THE ASTROPHYSICAL JOURNAL, 498 : 385È393, 1998 May 1
( 1998. The American Astronomical Society. All rights reserved. Printed in U.S.A.
HD 98800 : A UNIQUE STELLAR SYSTEM OF POSTÈT TAURI STARS1,2,3
DAVID R. SODERBLOM, JEREMY R. KING, LIONEL SIESS,4 KEITH S. NOLL, DIANE M. GILMORE, TODD J. HENRY,
EDMUND NELAN, CHRISTOPHER J. BURROWS, AND ROBERT A. BROWN
Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218 ; soderblom=stsci.edu, jking=stsci.edu, siess=stsci.edu, noll=stsci.edu,
dgilmore=stsci.edu, thenry=cfa.harvard.edu, nelan=stsci.edu, burrows=stsci.edu, rbrown=stsci.edu
M. A. C. PERRYMAN
Astrophysics Division, European Space Agency, ESTEC, Noordwijk 2200AG, Netherlands ; mperryma=astro.estec.esa.nl
G. FRITZ BENEDICT AND BARBARA J. MCARTHUR
McDonald Observatory, University of Texas, Austin, TX 78712 ; fritz=dorrit.as.utexas.edu, mca=barney.as.utexas.edu
OTTO G. FRANZ AND LAURENCE H. WASSERMAN
Lowell Observatory, 1400 West Mars Hill Road, Flagsta†, AZ 86001 ; ogf=lowell.edu, lhw=lowell.edu
BURTON F. JONES
University of California Observatories/Lick Observatory, Board of Studies in Astronomy and Astrophysics, University of California, Santa Cruz, CA 95064 ;
jones=ucolick.org
AND
DAVID W. LATHAM, GUILLERMO TORRES, AND ROBERT P. STEFANIK
Harvard-Smithsonian Center for Astrophysics, 60 Garden Street, Cambridge, MA 02138 ; latham=cfa.harvard.edu, torres=cfa.harvard.edu
Received 1997 August 12 ; accepted 1997 December 11
ABSTRACT
HD 98800 is a system of four stars, and it has a large infrared excess that is thought to be due to a
dust disk within the system. In this paper we present new astrometric observations made with Hipparcos,
as well as photometry from Hubble Space Telescope WFPC2 images. Combining these observations and
reanalyzing previous work allow us to estimate the age and masses of the stars in the system. Uncertainty in these ages and masses results from uncertainty in the temperatures of the stars and any
reddening they may have. We Ðnd that HD 98800 is most probably about 10 Myr old, although it may
be as young as 5 Myr or as old as 20 Myr. The stars in HD 98800 appear to have metallicities that are
about solar. An age of 10 Myr means that HD 98800 is a member of the postÈT Tauri class of objects,
and we argue that the stars in HD 98800 can help us understand why postÈT Tauris have been so
elusive. HD 98800 may have formed in the Centaurus star-forming region, but it is extraordinary in
being so young and yet so far from where it was born.
Subject headings : binaries : visual È circumstellar matter È stars : evolution È
stars : individual (HD 98800) È stars : preÈmain-sequence
1. HD 98800
AS A STELLAR SYSTEM
98800B, which are presently separated by about 0A. 8. Both
visible components are spectroscopic binaries, and we will
label these four stars hierarchically as Aa, Ab, Ba, and Bb.
The orbital and spectroscopic properties of these stars were
discussed in Torres et al. (1995) and Soderblom et al. (1996 ;
hereafter Paper I), respectively. To summarize, the Aa ] Ab
pair has an orbit with a period of 262 days and an eccentricity of 0.484, while the Ba ] Bb pair has a period of 315 days
and an eccentricity of 0.781 (Torres et al. 1995). The visual
pairÏs orbit is estimated to have a period of at least 105 yr
with an eccentricity of about 0.993 ; in other words, the
visual pair is marginally bound and we are seeing it near
periastron. Three of the stars appear in a Keck HIRES
spectrum, and the presence of abundant lithium in all three
(Paper I) argues that they are physically related objects, not
a chance superposition on the sky. All three stars rotate
slowly and show modest levels of chromospheric activity
(Paper I).
HD 98800 stands out among K stars because of its large
infrared excess. Sylvester et al. (1996) determined a fractional excess IR luminosity (i.e., the IR excess divided by the
stellar bolometric Ñux) of 0.084, but the actual value may be
twice that or more if the dust giving rise to this excess is
around one star or the other, not both. HD 98800 is now
well observed, with photometric data available from the
Zuckerman & Becklin (1993) have noted that ““ In many
ways, HD 98800 is the most unusual source in the IRAS
catalogs.ÏÏ We concur, and, as we will discuss here, HD
98800 may also be exemplary and instructive as an example
of a preÈmain-sequence (PMS) star found far from any
region of star formation. As we will show, the HD 98800
system is probably about 10 Myr old, meaning that it falls
into the category of postÈT Tauri stars (PTTs ; see Herbig
1978). The dearth of PTTs remains a fundamental problem
in stellar evolution, and we will argue that the stars in the
HD 98800 system demonstrate why PTTs may have been so
difficult to Ðnd.
First some nomenclature : what we call ““ HD 98800 ÏÏ consists of two visible objects, designated HD 98800A and HD
1 Based on observations obtained with the NASA/ESA Hubble Space
Telescope, obtained at the Space Telescope Science Institute, which is operated by the Association of Universities for Research in Astronomy, Inc.,
under contract with the National Aeronautics and Space Administration.
2 Based on data from the ESA Hipparcos astrometry satellite.
3 Based on observations obtained at the W. M. Keck Observatory,
which is operated jointly by the University of California and the California
Institute of Technology.
4 Laboratoire dÏAstrophysique de lÏObservatoire de Grenoble, Universite Joseph Fournier, BP53, F-38041, Grenoble Cedex, France.
385
386
SODERBLOM ET AL.
optical (Gregorio-Hetem et al. 1992), infrared (Garcia-Lario
et al. 1990 ; Zuckerman & Becklin 1993), and millimeterwave regimes (Sylvester et al. 1996). All these observations
do not resolve HD 98800 into its components and instead
treat it as a single object.
In an analysis of these data, Sylvester & Skinner (1996)
applied models to determine that HD 98800 has large (D1
mm) dust grains (needed to reproduce the millimeter-wave
Ñux). The nominal outer radius of the dust disk determined
from their best-Ðtting model is very large, but they can
accommodate an outer radius for the dust disk of as little as
25 AU, which means the dust can Ðt within the distance of
closest approach determined by Torres et al. (1995). The
estimated disk mass of Sylvester & Skinner (1996) is
4 ] 10~7 M .
_
These analyses account for some of the observations of
this system, but they leave more fundamental questions
unanswered : How old are the stars in HD 98800 ? If they
are main-sequence stars, as their luminosity classiÐcation
would suggest, how is it that a dust disk has survived for so
long when we anticipate that a typical disk lifetime is only
D107 yr (Skrutskie et al. 1990) ? Around which spectroscopic binary (A or B) is the dust disk, or do both pairs have
disks ? If HD 98800 is a PMS system, how did it get so far
from a star-forming region in such a short time ? Where was
it born ? Are there other young stars associated with it ?
What else can we learn from this system ?
We cannot answer all of these questions, but we address
them in this paper. Our essential datum is the parallax for
the system, which is critical for determining the luminosities
of these stars and hence their age. The parallax places HD
98800 at about 47 pc, unambiguously indicating that HD
98800 is a system of preÈmain-sequence objects. We have
also obtained other observations, including images with
Hubble Space Telescope (HST ). As we noted, HD 98800
appears to be a postÈT Tauri system, with some implications for that class of objects, but we cannot identify conclusively the origin of this extraordinary system.
2.
NEW OBSERVATIONS
2.1. Hipparcos
The recently released Hipparcos catalog (European Space
Agency 1997) provides details on the measurements made
with that satellite and the construction of the catalog. The
Hipparcos results for HD 98800 are listed in Table 1. The
(B[V ) and (V [I) (Cousins) colors in Table 1 are also from
that catalog. Hipparcos resolved Aa ] Ab from Ba ] Bb, as
shown by the parameters listed in Table 1.
Vol. 498
The Hipparcos observed magnitude versus time is shown
in Figure 1. HD 98800 is Ñagged as a variable object in the
output catalog because of the range of magnitudes seen. No
clear periodicity is seen in the variations. The Hipparcos
parallax of 21.43 mas leads directly to a distance of 47 pc.
We will determine luminosities and ages below, but this
distance implies that the stars in HD 98800 are too luminous to be main-sequence stars.
2.2. HST Fine Guidance Sensors
We used the Ðne guidance sensors (FGSs) on Hubble
Space Telescope during cycles 5 and 6 to measure the parallaxes and proper motions of the Aa ] Ab and Ba ] Bb
components separately. The parallax series, obtained with
FGS 3, occurred at three epochs, with two orbits of observations at each epoch and with each epoch separated from
the next by 6 months. This strategy has been used successfully on other targets to achieve parallaxes with precision of 1È2 mas (Benedict et al. 1995).
However, in this case the complexity and multiplicity of
the targets and the angular sensitivity of FGS 3 render the
data set insufficient. Both the POS and TRANS mode
observations appear to be perturbed by the duplicity of
both HD 98800A and HD 98800B. The expected orbital
period of each system is near 1 yr, which is nearly ideal for
contaminating parallax measurements that are taken D1 yr
apart. As a result, the FGS parallaxes for HD 98800 A 2and
B are in agreement with the Hipparcos value and show that
both stars are at the same distance, but the Hipparcos parallax is used here for determining the luminosities.
2.3. HST W FPC2
We undertook imaging observations with HST Ïs
WFPC2 camera in order to be able to determine separate
colors for A and B, and in the hopes of detecting evidence of
the dust disk around one star or the other. These observations are listed in Table 2. Stellar Ñuxes were measured
using the standard digiphot-apphot IRAF photometry
package. An aperture of 7 pixels was used, and the background was calculated inside an annulus located between
35 and 40 pixels from the star center, well outside both A
and B components. We corrected for Ñux outside the aperture using the Holtzman et al. (1995) correction and
TABLE 1
Hipparcos RESULTS FOR HD 98800 (HIP 55505)
Parameter
Value
R.A. (J1991.25) . . . . . . . . . . .
Decl. (J1991.25) . . . . . . . . . .
m .........................
n V(mas) . . . . . . . . . . . . . . . . . . . .
k (mas yr~1) . . . . . . . . . . . . .
ka (mas yr~1) . . . . . . . . . . . . .
d (mag) . . . . . . . . . . . . . . . . .
BT
V T (mag) . . . . . . . . . . . . . . . . .
(B[V ) (Tycho) . . . . . . . . . . .
(V [I) (Cousins) . . . . . . . . . .
Hp . . . . . . . . . . . . . . . . . . . . . . . . .
Position angle (deg) . . . . . .
o (arcsec) . . . . . . . . . . . . . . . . . .
11 22 05.34
[24 46 39.5
8.89
21.43 ^ 2.86
[85.45 ^ 1.89
[33.37 ^ 2.12
10.597 ^ 0.038
9.236 ^ 0.021
1.150 ^ 0.035
1.11 ^ 0.03
9.0217 ^ 0.0030
3
0.775 ^ 0.004
FIG. 1.ÈHipparcos Hp magnitude vs. day number. The range exceeds
that for nonvariable stars, which is approximately 0.01 mag. The points are
shown without error bars to reduce confusion, but the median and average
uncertainty of a single observation are both 0.017 mag for these 123 measurements.
No. 1, 1998
HD 98800
TABLE 2
WFPC2 OBSERVATIONS
Date
Exposure Time
(s)
Filter
Johnson
Magnitude
Aa ] Ab
1996
1996
1996
1996
1996
1996
Mar 3 . . . . . .
Mar 3 . . . . . .
Mar 3 . . . . . .
Jan 5 . . . . . . .
Jan 5 . . . . . . .
Jan 5 . . . . . . .
F439W
F555W
F953N
F439W
F555W
F953N
2.00
0.11
1.60
2.00
0.11
1.60
10.532
9.411
8.103
10.593
9.466
8.219
(B)
(V )
(I)
(B)
(V )
(I)
2.00
0.11
1.60
2.00
0.11
1.60
11.358
9.941
8.165
11.352
9.944
8.166
(B)
(V )
(I)
(B)
(V )
(I)
Ba ] Bb
1996
1996
1996
1996
1996
1996
Mar 3 . . . . . .
Mar 3 . . . . . .
Mar 3 . . . . . .
Jan 5 . . . . . . .
Jan 5 . . . . . . .
Jan 5 . . . . . . .
F439W
F555W
F953N
F439W
F555W
F953N
obtained resulting magnitudes in the STMAG system.
Small corrections are needed to convert the measured magnitudes from the F439W and F555W Ðlters to the Johnson
B and V bands ; These were performed using the IRAF
synphot task calcphot. We selected a K4V stellar spectrum
from the Bruzual-Persson-Gunn-Stryker spectrum atlas
(see Bushouse 1995), determined the correction (Johnson
B \ F439W ] 0.5005 mag ; Johnson V \ F555W [ 0.0072
387
mag), and Ðnally obtained the Johnson B and V magnitudes
for both components. The images obtained in 1996 January
are compromised photometrically because the stars fell near
a charge trap ; in 1996 March the stars fell on a good
portion of the CCD, and so the summary data are based on
the 1996 March observations. The combined V magnitude
we derive for A ] B is 8.89, exactly that measured on the
ground (Gregorio-Hetem et al. 1992). The combined B magnitude is 10.10, again consistent with the ground-based
combined magnitude of 10.14. Our longest wavelength
observations, in the narrowband F953N Ðlter, require a
large correction (Johnson I \ F953N [ 1.4735 mag) to be
compared to ground-based I magnitudes and should therefore be viewed with caution, but the magnitude seen in that
band is consistent with the (V [I) color listed with the Hipparcos observations.
We could see no evidence in the images for nonÈpointsource emission (Fig. 2). We conÐrmed this by determining
a normalized and azimuthally averaged radial proÐle for
star Aa ] Ab and then comparing that to the model pointspread function (PSF) computed with the software package
TinyTim (Krist 1993). This comparison is shown in Figure
2b. ProÐles for Ba ] Bb are not shown, but they had the
same point-source PSF.
2.4. Discussion and Summary of the Observations
Because HD 98800 is a multiple system, the Hipparcos
proper motion is a measure of the mean photocentric
N
2500
2500
PSF
Intensity
2000
1"
Aa
2000
1500
1500
1000
1000
500
500
0
0
0.00
0.05
0.10
0.15 0.00
0.05
0.10
0.15
Radius (arcsec)
FIG. 2a
FIG. 2b
FIG. 2.È(a) WFPC2 F555W image of the HD 98800 system, presented with a logarithmic stretch. (b) Radial proÐle of a PSF computed using the TinyTim
software package (crosses, left panel). A jitter of 15 mas was assumed in the computation. The solid curve is a cubic-spline Ðt to the plotted points. In the right
panel the observed points from the image obtained in 1996 March (F439W, diamonds) are shown as a function of radius from the center of the image. The
position of the center was determined with the IRAF imcenter task. The same cubic spline from the left panel is shown again, multiplied by a constant scale
factor. The excellent agreement between the computed and observed PSFs for the Aa ] Ab pair indicates that they are unresolved. A similar analysis for the
other Ðlters and for the Ba ] Bb pair also shows no evidence for extended structure.
388
SODERBLOM ET AL.
TABLE 3
SUMMARY OF OBSERVATIONS OF HD 98800
Parameter
Value
n (mas) . . . . . . . . . . . . . . . . . . . . . . . . .
Distance (pc) . . . . . . . . . . . . . . . . . . .
k (mas yr~1) . . . . . . . . . . . . . . . . . .
ka (mas yr~1) . . . . . . . . . . . . . . . . . .
d
Radial velocity (km s~1) . . . . . .
U (km s~1) . . . . . . . . . . . . . . . . . . . . .
V (km s~1) . . . . . . . . . . . . . . . . . . . . .
W (km s~1) . . . . . . . . . . . . . . . . . . . .
V
(mag) . . . . . . . . . . . . . . . . . . .
Aa`Ab
V
(mag) . . . . . . . . . . . . . . . . . .
Ba`Bb
(B[V )
(mag) . . . . . . . . . . . .
Aa`Ab
(B[V )
(mag) . . . . . . . . . . . . .
Ba`Bb
21.43 ^ 2.86
46.7 ^ 6.2
[89 ^ 3.1
[23 ^ 2.9
]12.75 ^ 0.1
[13.2 ^ 2.1
[19.7 ^ 1.3
[3.5 ^ 1.5
9.41
9.94
1.11
1.40
motion over the interval of the mission, and that may be
unrepresentative of the long-term motion of the systemÏs
center of mass (because of the missionÏs short time baseline
and the orbital motion of the stars). Therefore we use the
PPM (Bastian & RoŽser 1993) proper motions here. We
adopt the systemic radial velocity of Torres et al. (1995), and
use the procedure in Johnson & Soderblom (1987) to calculate the components of the space velocity vector for the
J2000 reference frame ; these are listed in Table 3.
For the present, we take the WFPC2 photometry to represent the true magnitudes and colors of HD 98800 A and
B. There may be circumstellar reddening present, but
Sylvester et al. (1996) argue that the dust grains in the HD
98800 system must be large (Z1 mm) in order to account
for the observed millimeter-wave Ñuxes, and such particles
will produce gray extinction.
3.
THE EVOLUTIONARY STATUS OF HD 98800
3.1. L ocation of the Stars in the H-R Diagram
The ages of these stars are determined by their temperatures and luminosities, both of which are inferred from
the photometry. At a distance of only 47 pc, HD 98800 is
unlikely to exhibit signiÐcant interstellar reddening, but its
infrared excess suggests that circumstellar reddening may
be present.
First, we consider the simplest case (case A), for which the
temperatures are derived taking the WFPC2 photometry at
face value and using BessellÏs (1979) color-temperature relation. We also assume that no reddening or extinction are
present. The T value for component Aa, assumed to be
eff
negligibly contaminated
by its companion, follows directly
from the photometry ; its absolute magnitude follows
directly from the photometry and parallax. The relative T
eff
values for the Ba and Bb components with respect to the Aa
component were assumed to be identical to those employed
in Paper I. The (relatively age-independent) di†erences in
M for the Ba and Bb T values were estimated from our
V
eff
stellar
models (Siess, Forestini,
& Dougados 1997), and
used to deconvolve the single WFPC2 V magnitude for the
B component into individual values for Ba and Bb.
Figure 3 shows the locations in an H-R diagram of stars
Aa, Ba, and Bb for the three cases we describe here. (Note
that the position of star Bb is shown using the temperature
estimated in several ways, but that the luminosity follows
from assuming that Bb is the same age as Aa and Ba, and so
is not an independent point.) PreÈmain-sequence tracks are
from Siess et al. (1997), as is the conversion table between
Vol. 498
luminosity and magnitude. The evolutionary paths are
computed for a solar metallicity (Z \ 0.02,) with relative
abundances from Anders & Grevesse (1989). These stars are
located near the bottom of the Hayashi track, a radiative
core has already developed, and Li is now burning in their
convective envelopes. The stellar gravity is very similar for
all components : log g B 4.25. From the position of the stars
in the H-R diagram, we estimate an age for the system of
about 7 Myr. The masses of the three components are
between 0.8 and 1.0 M for Aa, 0.75 and 0.85 M for Ba
_
_
and 0.45 and 0.6 M for Bb (Table 4).
_
As a second scenario (case B), we use the T values of
eff
Paper I, which were estimated from the Keck HIRES spectrum, together with the same absolute magnitudes of case A
(Fig. 3a) ; these same magnitudes are justiÐed for the Ba and
Bb components because, while the T values are slightly
eff and they predict
di†erent, the relative values are the same
the same M di†erence used to deconvolve the photometry
V
of the B component.
Note that in this case the derived ages
of Aa and Ba agree with each other, but at 12 Myr they are
somewhat older than for case A. If the case B T values are
eff
correct and the color-temperature relation is correct,
then
the observed colors would suggest some reddening : about
0.06 mag in B[V for Aa and 0.08 mag for Ba.
Another possibility, case C, was motivated by investigation of the Li lines described in the next section. An extensive series of syntheses in the 6708 AŽ Li I spectrum region
lead us to believe that the relative Ñuxes of the Aa, Ba, and
Bb components in this regime are near 50%, 40%, and
10% ; these are close to the values of Paper I. With initial
guesses at the T values (essentially those of case A), we
eff
solved for the relative
radii of the components. These then
yielded predicted Ñuxes at V , which were compared to our
WFPC2 photometry. Since initial di†erences are found, we
then determined what extinction of the Ba and Bb components would be necessary to explain the di†erence in the
Li-based predicted Ñuxes and the observed Ñuxes. At the
same time, the predicted Ñuxes allowed us to deconvolve the
combined WFPC2 V magnitudes of Ba and Bb (with model
calculations as described above) and to reÐne the initial T
eff
guesses. The procedure is then iterated. We found that an
extinction of A B 0.44 mag for Ba and Bb could explain
V
the predicted Li-based
and observed HST -based V Ñux differences.
TABLE 4
EVOLUTIONARY PARAMETERS OF HD 98800
T
eff
(K)
Component
Case A T
Aa . . . . . . . . . .
Ba . . . . . . . . . .
Bb . . . . . . . . . .
4350
4100
3550
Case B T
Aa . . . . . . . . . .
Ba . . . . . . . . . .
Bb . . . . . . . . . .
Case C T
Aa . . . . . . . . . .
Ba . . . . . . . . . .
Bb . . . . . . . . . .
Age
(Myr)
M/M
V
_
from WFPC2 Photometry
M
eff
6.06
6.79
8.5 :
from
eff
4500
4250
3700
7
7
...
log N(Li)
0.95
0.75
0.45
3.1
2.0
3:
Paper I, No Reddening Correction
6.06
6.79
8.5 :
12
12
...
1.00
0.85
0.6
3.1
2.2
1.8
from Paper I, Ba ] Bb Reddening Corrected
eff
4350
6.06
7
0.95
3.1
4250
6.36
7
0.85
2.3
3700
8.1 :
...
0.5
3:
No. 1, 1998
HD 98800
389
FIG. 3.ÈH-R diagram and location of the di†erent components of HD 98800. Each mass track (solid lines) is labeled in solar units. From top to bottom,
the isochrones (dotted lines) correspond to ages of 1, 1.5, 3, 5, 10 (heavy dotted line), 20, and 40 Myr. Uncertainties are approximately 100 K in T and 0.3 mag
eff position is
in M . L eft panel : Cases A (circles) and B (triangles). Right panel : Case C. The points for HD 98800Bb are shown as open points because their
V from the temperature for a particular case assuming Bb has the same age as Aa or Ba. For any one case the stars are in the order Aa, Ba, and Bb from
inferred
hottest to coolest.
Assuming a standard interstellar reddening law of A \
V
3.1E
(which may be questionable despite the possibility
(B~V)
that gray extinction may be present ; Sylvester & Skinner
1996), we arrive at Ðnal T estimates for case C of 4350,
4250, and 3700 K for theeffAa, Ba, and Bb components,
respectively. The Ðnal deconvolved V magnitudes and Hipparcos parallax then resulted in M \ 6.06, 6.36, and 8.1 for
Aa, Ba, and Bb, respectively. ThisVprocedure yields consistent age estimates of 7 Myr for both the Aa and Ba components.
No one of these three cases is a perfect match, but case C
is our best guess for the state of the HD 98800 system.
3.2. Age and L ithium Abundance
In Paper I it was noted that the large Li abundances of
the HD 98800 components were consistent with youth, but
that an exact age could not be estimated because of the
difficulty of understanding Li depletion in solar-type stars.
These large abundances are a robust conclusion that can be
affirmed from simple visual comparison of Li j6708 to the
nearby Ca 6717 AŽ line. Figure 4 of Paper I shows that the Li
lines exceed the Ca lines in depth (at least for the Aa and Ba
components). In our experience of observing Li in a number
of Population I F, G, and K dwarfs, this is a rare circumstance, except in young, Li-rich stars such as some found in
the Pleiades. We also compared the relative Li/Ca line
strengths to those of cool Pleiades dwarfs, as shown in the
Keck HIRES spectra in Figure 1 of Jones et al. (1996),
conÐrming that the Li abundances in the stars of HD 98800
are larger than similar Pleiades stars.
As noted above, we also examined whether these features
could yield additional information on the relative Ñux contributions of the HD 98800 components. One might think
that seeking to constrain the abundances and relative Ñuxes
for all three components from just three features is fruitless,
but this is not exactly true. In particular, in the regime of
large Li abundances we are dealing with here, the intrinsic
(undiluted) features are very strongÈof near-zero continuum depth. Thus if one errs by underestimating the Ñux
contribution of a given component, one cannot simply
increase the Li abundance to reproduce the observed line
depth. Instead, such alterations tend to have substantial
e†ects on the line wings, and not the line depths. So while
there is some degeneracy, the observed Li line depths tend
to be sensitive to the relative Ñux contributions, while the
line proÐle shape away from line center provides valuable
constraints on the Li abundances.
Spectrum synthesis of the 6708 AŽ region was conducted
with an updated version of the LTE analysis package
MOOG (Sneden 1973), using the model atmosphere grids of
Kurucz (1992, private communication). The line list was
taken from the recent work of King et al. (1997). Many
combinations of trial T values for the three stars were
utilized. These includedeffthe favored values from Paper I,
values adjusted from Paper I given the new HST photometry, and a range of a few hundred degrees in these relative
T values of the three components visible in the spectrum
eff could account for extinction of one or more of the
(this
components, for example). Arbitrarily varying the Li abundances for all these di†erent T values, we sought to
eff Figure 4 of Paper I. In
improve the synthetic Ðt shown in
particular, we desired to Ðt better the maximum Ñux seen in
the red (blue) edge of the Ba (Aa) component without
destroying agreement in the line core or making the line
wings too broad. This led to a slightly increased Ñux contribution for Ba at 6707 AŽ , compared to that used in Paper I.
390
SODERBLOM ET AL.
As a result, it then became challenging to Ðt the core and
red wing of the Aa component.
We found consistently that the relative Ñux contributions
in the 6708 AŽ region were within a few percent of 50%, 40%,
and 10% for the Aa, Ba, and Bb components, respectively.
These are close to the values of 50%, 35%, and 15% from
Paper I, and they produced good agreement between the
observed and synthetic Li proÐles. The Li results are summarized in Table 4. These abundances have each been
lowered by approximately 0.1 dex to correct for non-LTE
e†ects (the actual corrections are [0.02 dex for star Ba in
cases B and C, [0.07 dex for Ba in case A, [0.08 dex for
Bb in case A, [0.09 dex for star Aa in all cases and Bb in
case C, and [0.16 dex for Bb in case B). The LTE abundances for case B were taken from Paper I. Those for cases
A and C are those derived here using the above Ñux ratios.
Unfortunately, none of the several satisfactory Ñux
contribution/parameter combinations we found (including
case C above) resulted in very good simultaneous agreement with the photometric Ñux ratios, relative model radii,
and the Bb/Ba mass ratio (0.834) from the orbital solution
of Torres et al. (1995).
The main results from this are as follows. First, the plausible combinations we explored did suggest larger Li abundances in Aa than in Ba. Abundances for the former
component ranged from 2.7 to 3.3 [on the scale of log
N(H) \ 12], while abundances for the latter component
ranged from 2.1 to 2.8. Second, the case C scenario indicated that the Li abundance of the cooler Bb could be larger
than that of the hotter Ba component. While open cluster
and Ðeld star observations have repeatedly shown a pattern
of declining Li abundance with declining T (decreasing
eff stars, as we
mass), this behavior of Li increasing for cooler
see here for HD 98800, is not unexpected. Cooler stars have
deeper convection zones, but at some T dependent on (at
eff and densities at
least) age and metallicity, the temperatures
the base of the convection zones are simply too low to
deplete as much Li in as much time as a slightly hotter star
with a shallower convection zone. This behavior is analogous to that which makes the presence of Li useful in conÐrming the brown dwarf status of exceedingly cool (but
presumably signiÐcantly older than HD 98800) objects in
the Ðeld and young clusters. Indeed, standard evolutionary
models we constructed predict this upturn in Li abundance
over the range 4200È3900 K for ages of 5È12.5 Myr.
A comparison of our abundances to stellar models is
presented in Figure 4, which shows surface Li abundance as
a function of T for various ages. In our computations, the
eff is taken to be log N(Li) \ 3.31, the meteinitial Li content
0 we can see, for a
oritic value (Anders & Grevesse 1989). As
given case the age of the system is strongly constrained. For
case B, for example, we derive an age between 10 and 12
Myr. This constraint arises from the component Bb, less
massive and still burning its Li efficiently at the bottom of
its deep convective envelope. For more massive stars (Aa
and Ba), the temperature at the base of the convective
envelope is lower, and consequently Li surface abundance is
depleted more slowly.
3.3. Conclusions about Age
The location of the stars of the HD 98800 system in the
H-R diagram unambiguously indicates that they are preÈ
main-sequence stars, and that they are at the bottom of
their Hayashi tracks. A comparison of Li abundances to
Vol. 498
FIG. 4.ÈEvolution of the surface Li abundance as a function of T .
eff
Isochrones are equally spaced in time by 1 Myr, starting with 3 Myr at the
top to 17 Myr at the bottom. The heavy line corresponds to 10 Myr.
Uncertainties are approximately 0.3 dex in the Li abundance and 100 K in
T . The squares correspond to case A, the solid circles to case B, and the
eff circles to case C.
open
stellar evolution models constrains the age of HD 98800 to
D10 Myr old, with an uncertainty due to modeling of about
3 Myr. Thus from its evolutionary status HD 98800 presents the characteristics of postÈT Tauri stars, ending their
fully convective phase and joining a radiative path toward
the main sequence.
4.
METAL ABUNDANCES
We reanalyzed the Keck HIRES spectrum of the HD
98800 system described in Paper I in an attempt to derive
detailed abundances of several metallic elements. In the end
we were unable to derive satisfactory abundances. This is
due not to complications such as the choppy continuum,
frequent severe blending, and the necessity of correcting for
the e†ects of Ñux dilution ; we believe these were handled
reliably. Rather, we encountered a wavelength dependence,
described below, in the relative line strengths of neutral
metal features over only 1000 AŽ . We realized later that this
e†ect is probably purely observational and arises from
atmospheric refraction but that it may compromise our
analysis to some extent.
Abundances were determined from measured line
strengths using the MOOG package. Lines were checked
for ““ cleanliness ÏÏ by direct inspection, inspection of the
Kurucz et al. (1984) solar Ñux atlas, and by inspection of
synthetic spectra including features within several angstroms of the line in question using the Kurucz CD-ROM
line lists. The same lines retained for analysis were measured in our high signal-to-noise HIRES solar proxy spectrum of the daytime sky and analyzed in the same fashion
so that detailed accuracy of the absolute gf values is not
very important and so that assumed solar abundances need
No. 1, 1998
HD 98800
not be adopted. IdentiÐcations and atomic data came from
the list of Thevenin (1990).
The resulting line-by-line abundance di†erences between
Aa and Ba show a marked trend with wavelength ; this is
approximately 0.20 dex at 6500 AŽ , and increases to about
0.30 dex near 7700 AŽ . Further investigation showed that
this resulted from a curious trend in the Aa component
alone. We thus conducted similar analyses of two cool
single Pleiades dwarfs using the HIRES spectra discussed in
Jones et al. (1996). These Pleiades dwarfs are of similar
temperature to the stars in HD 98800, and their spectra
were obtained at the same time with the same equipment as
the HD 98800 spectrum. No such trends are seen in the
Pleiades dwarfs, indicating there is no fundamental Ñaw in
the analysis procedure or the software.
Instead, the cause apparently can be traced directly to the
relative line strengths. Inspection of features of the same
species having similar excitation potential in the blue and
red ends of our spectra show di†erences ; as suspected, the
Aa componentÏs features appear shallower relative to Ba in
the red. In hindsight, this e†ect might be deduced from
Figure 3 of Paper I. The top panel there shows the Ni I
j7522 and j7524 lines. It can be seen that the Ba component
exhibits deeper lines than the Aa component. On the other
hand, from much bluer spectra at 5180 AŽ , Torres et al.
(1995) identify Aa as the primary component. Such a conclusion from Figure 3 of Paper I is not at all obvious. The
two works are, however, using the same nomenclature. This
was veriÐed by noting that the Ba and Bb components in
Paper I showed the radial velocity shift predicted by the
orbital solution of Torres et al. (1995).
Relative line depth ratios of features of the same species
having nearly identical excitation potential were measured
for the Aa and Ba components. These were then compared
to the ratios demonstrated by the same features in the
HIRES spectra of cool Pleiades dwarfs. This exercise again
indicated that the Aa component was the source of the
discrepancy. This was further veriÐed by comparing the
relative line depth ratios to ones synthesized with a range of
parameters and Ñux ratios.
However, we note that our composite spectrum of HD
98800 was obtained at a large zenith distance, and that two
stars separated by about 0A. 8 were being observed with an
aperture 0A. 86 wide. Guiding errors could lead to a systematic o†set in the ratio of the light from Ba relative to Aa, but
atmospheric refraction could account for a di†erential e†ect
with wavelength. We suspect that this is the origin of this
apparent discrepancy in the abundances in Aa relative to
Ba. Thus we cannot yet present reliable detailed abundances. At present, though, experimentation with a wide
variety of parameter combinations suggests to us that the
abundance of all elements considered is within 0.2 dex of the
solar values, which is typical for most stars of the solar
neighborhood. However, we believe that we are not far
amiss for two reasons. First, the guiding was done with a
red Ðlter, so that the center of light of the A ] B pair should
have been well centered on the aperture. This means that
the infrared wavelengths may yield unreliable abundances
but that the shorter wavelengths should be all right. Second,
as we noted above the strength of the Li doublet allows an
analysis to be made that is nearly independent of knowledge of the relative Ñuxes. We remain conÐdent in our
conclusions about Li and the roughly solar estimate for
[Fe/H].
391
5.
DISCUSSION
5.1. T he Kinematic Origin of HD 98800
The HD 98800 system is very young, yet it appears far
from any obvious region of star formation. Where was it
born ? Using the proper motion, parallax, and radial velocity of HD 98800, we calculated the space motion (see Table
3). After correcting for a basic solar motion of (U, V ,
W ) \ ([9, ]11, ]6) km s~1, we determined that the position of HD 98800 was at (X, Y , Z) \ ([30, ]59, ]15) pc
relative to the Sun 10 Myr ago, in a coordinate system in
which the x-axis points in the Galactic anticenter direction,
the y-axis is in the direction of Galactic rotation, and the
z-axis toward the north Galactic pole. For this calculation,
we have assumed rectilinear motion, which is probably a
fair assumption for the X and Y positions but may not be
for Z. Given the low velocity dispersion of young objects,
we looked at regions currently within 400 pc of this position
and with measured proper motions and radial velocities
and asked where they were 10 to 20 Myr ago. The only
region that appears to be a likely point of origin for HD
98800 is the Scorpio-Centaurus (Sco OB2) complex, which
consists of upper Scorpius, Upper Centaurus-Lupus, and
Lower Centaurus-Crux.
Sco-Cen is a large, complex region of star formation that
is part of the Gould Belt. It stretches across D70¡ of sky,
corresponding to a breadth of D200 pc and a thickness of
60 pc at its distance of D165 pc. Proper motions and radial
velocities from Bertiau (1958) were used to project three
centers within the complex backward in time. The results of
this projection are shown in Figure 5 ; the points of the
arrows show the current locations of HD 98800 and parts of
the Sco-Cen complex. The plus sign at the center is the
current location of the Sun. The bases of the arrows show
the locations 20 Myr ago (with respect to the LSR), and the
large dots show the positions 10 Myr ago.
Given the large size of Sco-Cen, Figure 5 shows that in
the past HD 98800 was within or close to parts of the
complex. But when was HD 98800 closest to these regions,
and how far apart were the regions and HD 98800 at that
time ? Looking only at the distances projected on the Galactic plane, the closet approach to upper Scorpius was 22 Myr
ago at a distance of 19 pc, and to upper Centaurus-Lupus
15 Myr ago at a distance of 30 pc. Given the size of the
complex and the uncertainty of the calculation, we conclude
that it is reasonable that HD 98800 was within the complex
between 10 and 30 Myr ago. It also appears more likely that
HD 98800 is associated with one of the Centaurus regions
than it is with upper Sco.
The expansion age for various parts of the Sco-Cen
complex is D20 Myr, and the evolutionary age for upper
Cen-Lupus and lower Cen-Crux is D10 Myr (Blaauw 1964 ;
de Geus, de Zeeuw, & Lub 1989). Thus the time of closest
approach is consistent with the age of parts of the complex.
If this is indeed the origin of HD 98800, then it puts a
constraint on the age to be less than D20 Myr and makes
an age of 10 Myr more likely than 7 Myr, although upper
Sco may be as young as 5 Myr (de Geus et al. 1989).
HD 98800 is qualitatively similar to some G and K stars
that may belong to Sco-Cen. Park & Finley (1996) reported
on some objects in Crux found with ROSAT . Feigelson &
Lawson (1997) obtained low-resolution optical spectra of
them, and they have Ha emission strengths very similar to
HD 98800, although their Li lines appear somewhat
392
SODERBLOM ET AL.
FIG. 5.ÈLocations of HD 98800 and the Sco-Cen complex now and in
the past with respect to the LSR. The heads of the arrows show the current
locations, the bases of the arrows show the locations 20 Myr ago, and the
large dots show the positions 10 Myr ago. The plus sign is the current
position of the Sun.
weaker. HD 98800 is also similar in Ha emission strength
and Li abundance to some stars studied by Walter et al.
(1994) in upper Scorpius.
5.2. W hereÏs the Dust ?
We see no evidence in our WFPC2 images for a b
Pictoris-like dust disk, so we cannot yet pinpoint where the
infrared excess is being produced. The unusual orbit of the
visual pair suggests that the extraordinary IR excess of the
HD 98800 system is related to the fact that the Ba ] Bb
pair is near periastron relative to Aa ] Ab. Perhaps the
disk is being warped by the proximity of the other more
massive star, leading to more absorption of the central
starÏs light. Another possibility, mentioned in Paper I, is
that one of the stars, say Aa, illuminates a disk that is in the
Ba ] Bb system. The solid angle the disk subtends, as seen
from the companion star, is very large, allowing much of the
companionÏs light to heat the disk. This gets around the
usual problems of geometry when a star illuminates
material around itself, but if the optical depth of a disk is
very low in the direction parallel to its symmetry axis
(because of a low Ðlling factor), then this scenario may not
work. Models of this type could be calculated to see
whether this possibility is reasonable.
We suspect that the dust disk may be around the
Ba ] Bb pair because those stars look redder than we
expected, based on estimates of temperature from the composite spectrum, but that is only a hunch (and conÑicts with
large grains giving rise to gray extinction). near-infrared
camera and multiobject spectrometer (NICMOS) observations of this system should resolve this question.
5.3. HD 98800 as a System of PostÈT Tauri Stars
Their locations in the H-R diagram and their likely
relationship to Sco-Cen argues for the HD 98800 system
being close to 10 Myr old. There are other stars known to
be about this age, but nearly all of them are in or near
star-forming regions, if only because that is where searches
Vol. 498
have been made. Star-forming regions typically contain T
Tauri stars that are about 1 to 3 Myr old.
As Herbig (1978) pointed out, if stars have been formed at
a constant rate there should be an order of magnitude more
postÈT Tauris (PTTs) than there are T Tauris (TTs) because
of the timescales for evolution on the Hayashi track relative
to the remainder of the PMS evolution. Herbig tried several
ways to search for PTTs, without success (with the possible
exception of FK Ser ; Herbig 1973), and subsequent
attempts have turned up only a few. More recently, X-ray
imaging has been used as a means to Ðnd PTTs. We know
TTs are bright in X-rays, with log L /L B [3.8, and that
X bol
zero-age main sequence (ZAMS) stars
are relatively bright
too : log L /L B [4.2 (Bricen8 o et al. 1997). That suggests
X bol
PTTs should be found at intermediate levels of X-ray emission, and those levels are well within the reach of, e.g.,
ROSAT .
ROSAT has been used to observe the environs of starforming regions to Ðnd PTTs, and candidate objects have
been found, but their interpretation is not clear. NeuhaŽuser
et al. (1997) and Magazzù et al. (1997), for example, report
ROSAT observations of the Taurus-Auriga region and the
surrounding area, about 580 square degrees in all. They list
about 110 new T Tauri and PTT stars. Bricen8 o et al. (1997),
however, note that foreground ZAMS stars can account for
these detections both in number and Ñux level, which would
indicate that few or no PTTs were in fact in the Tau-Aur
region. Wichmann et al. (1997) have examined such stars in
the Lupus region and Ðnd that their Li-rich stars fall within
the Gould Belt, suggesting they are bona Ðde T Tauris, not
ZAMS stars.
There are several resolutions to this problem. First, as
noted by Palla & Galli (1997), perhaps the ““ problem ÏÏ is
nonexistent because stars with ages of D10 Myr are simply
not present near the Sun due to the episodic, as opposed to
steady, nature of star formation. A second solution would
be that PTTs exist but that they are far from star-forming
regions and so dispersed that they are difficult to Ðnd. A
third possibility is that PTTs exist and are abundant, but
they have remained invisible to us because our searches are
inappropriate.
The Ðrst possibility is beyond the scope of this paper, of
course. The third is partly plausible if we consider the
optical spectrum of the HD 98800 system. As we noted in
Paper I, all three stars have very low v sin i values and
correspondingly low levels of chromospheric activity, as
seen in the Ca II infrared triplet lines. This seems surprising
for such young stars, and especially so for star Bb, since it is
about half a solar mass. Other very young stars of similar
mass produce strong activity, enough so that they would
contribute signiÐcantly to an emission spike at Ha, even if
their Ñux contribution to the system would be modest. Thus
if these stars in HD 98800 are typical for an age of 10
MyrÈand there is no reason to believe they are not
typicalÈPTTs cannot be found through the methods that
have been tried, such as objective prism surveys.
X-rays are another matter. Kastner et al. (1997) report
ROSAT data for HD 98800, and the X-ray Ñux is more than
enough to make this system detectable even if it were moved
as far away as Tau-Aur. They report log L /L \ [3.4 for
X TTs.
bol Thus the
HD 98800, which is even brighter than most
X-ray searches for PTTs should Ðnd them, at least if they
are within the Ðelds examined.
That brings us back to the second possibility, namely,
No. 1, 1998
HD 98800
that PTTs are really there but theyÏre far removed from
regions of star formation, making them difficult to Ðnd. The
mere existence of a system like HD 98800 shows that this is
possible. Moreover, the fact that the visual orbit has high
eccentricity and a very long period means that Aa ] Ab is
only weakly bound to Ba ] Bb. That means that HD 98800
cannot have found its way so far from where it was born
through an unusual and energetic gravitational encounter
that spit it out. Instead, HD 98800 had to leave its nursery
through some gentle means, and that suggests it is more
typical than not.
HD 98800 is about 150 pc from Sco-Cen (Fig. 5), so its
velocity relative to Sco-Cen must be about 15 km s~1 for it
to have gotten to where it is within 10 Myr. Feigelson (1996)
suggests that dispersal of TTs takes place through two
mechanisms. The Ðrst is just due to the thermal velocity
within a molecular cloud. But that is typically only about 1
km s~1, much too little to account for HD 98800. FeigelsonÏs second mechanism is formation of stars in short-lived,
rapidly moving cloudlets, which seems like a plausible
origin for the HD 98800 system. Taurus-Auriga is not
known to have clouds within it moving away as fast as 15
km s~1. HD 98800 may be part of a group of very young
stars (Kastner et al. 1997), another of which is TW Hya.
Hipparcos measured a parallax for TW Hya that leads to a
space motion consistent with that of HD 98800 if TW HyaÏs
radial velocity is about ]15 km s~1. Sco-Cen is a signiÐcantly larger star-forming region than Tau-Aur, so perhaps
HD 98800 and its cohort are on the high-velocity end of the
distribution of a large population.
To identify other possible members of an HD 98800
cohort, we calculated space motions for the IR excess stars
in Table 1 of Sylvester et al. (1996). Many of those stars do
not have reported radial velocities, so we began by calculating U, V , and W for a range of radial velocities from [50
to ]50 km s~1 to see whether these lines converged in
velocity space. Several did, and follow-up checks showed
that one star, HD 23680, has a reported radial velocity
393
(]16.5 km s~1, Barbier-Brossat & Petit 1990) that would
give it a space motion indistinguishable from that of HD
98800. However, HD 23680 is nearly 200 pc away from the
Sun (its Hipparcos parallax is 5.54 mas) in a very di†erent
direction from HD 98800, and so it seems unlikely to have
had a common origin.
How many more stars or systems like HD 98800 might
there be ? Some signiÐcant number needs to be identiÐed if
we are to assess the prevalence of PTTs in the solar neighborhood. A few other examples of HD 98800Èlike objects
may exist. First, there is FK Ser (BD [10¡4662), HerbigÏs
prototype PTT, which, curiously, is a visual binary. Unfortunately, its Hipparcos parallax is too imprecise (9.42 ^ 6.17
mas) to allow its age to be determined from its luminosity.
Also, FK Ser has Ha emission more like that of a traditional
TT than the Ha seen in HD 98800. Jeffries et al. (1996)
obtained optical spectra of a double system detected by
ROSAT , 2RE J0241[525, and suggest that it too may be a
PTT. It is again unfortunate, but this latter binary was not
in the Hipparcos Input Catalog and so no parallax exists.
But it is highly suggestive that binaries are common, not
rare, among stars indicted as PTTs ; whatever process puts
such stars into the Ðeld is apparently not disruptive of their
environs.
The assistance of R. Riebau is gratefully acknowledged.
This work was supported in part by a grant from NASA.
Portions of these observations were made at the W. M.
Keck Observatory. The W. M. Keck Observatory is operated as a scientiÐc partnership between the California Institute of Technology and the University of California. It was
made possible by the generous Ðnancial support of the
W. M. Keck Foundation. L. S. acknowledges support from
the French Ministry of Foreign A†airs (Bourse Lavoisier),
and thanks STScI for its hospitality. The comments of an
anonymous referee were very helpful in improving the discussion.
REFERENCES
Anders, E., & Grevesse, N. 1989, Geochim. Cosmochim. Acta, 53, 197
Kastner, J. H., Zuckerman, B., Weintraub, D. A., & Forveille, T. 1997,
Barbier-Brossat, M., & Petit, M. 1990, Catalogue Bibliographique de
Science, 277, 67
Vitesses Radiales Stellaires (Marseille : Obs. de Marseille)
King, J. R., Deliyannis, C. P., Hiltgen, D. D., Stephens, A., Cunha, K., &
Bastian, U., & RoŽser, S. 1993, PPM Star Catalogue (Heidelberg : Astr.
Boesgaard, A. M. 1997, AJ, 113, 1871
Rechen-Inst.)
Krist, J. 1993, in ASP Conf. Ser. 52, Astronomical Data Analysis Software
Benedict, G. F., et al. 1995, in Astronomical and Astrophysical Objectives
and Systems II, ed. R. J. Hanisch, R. J. V. Brissenden, & J. Barnes (San
of Sub-Milliarcsecond Optical Astrometry, ed. E. Hoeg & P. K. SeidelFrancisco : ASP), 536
mann (Dordrecht : Kluwer), 89
Kurucz, R. L., Furenlid, I., Brault, J., & Testerman, L. 1984, Solar Flux
Bertiau, F. C. 1958, ApJ, 128, 533
Atlas from 296 to 1300 nm, National Solar Observatory Atlas No. 1
Bessell, M. S. 1979, PASP, 91, 383
Magazzù, A., Mart• n, E. L., Sterzik, M. F., NeuhaŽuser, R., Covino, E., &
Blaauw, A. 1964, ARA&A, 2, 213
Alcala, J. M. 1997, A&AS, 124, 449
Bricen8 o, C., Hartmann, L. W., Stau†er, J. R., Gagne, M., Stern, R. A., &
NeuhaŽuser, R., Torres, G., Sterzik, M. F., & Randich, S. 1997, A&A, 325,
Caillault, J.-P. 1997, AJ, 113, 740
647
Bushouse, H. 1995, Synphot Users Guide (3d ed. ; Baltimore : STScI)
Palla, F., & Galli, D. 1997, ApJ, 476, L35
de Geus, E. J., de Zeeuw, P. T., & Lub, J. 1989, A&A, 216, 44
Park, S., & Finley, J. P. 1996, AJ, 112, 693
European Space Agency. 1997, The Hipparcos and Tycho Catalogues
Siess, L., Forestini, M., & Dougados, C. 1997, A&A, 324, 556
(Paris : ESA), SP-1200
Skrutskie, M. F., Dutkevich, D., Strom, S. E., Edwards, S., Strom, K. M., &
Feigelson, E. D. 1996, ApJ, 468, 306
Shure, M. A. 1990, AJ, 99, 1187
Feigelson, E. D., & Lawson, W. A. 1997, AJ, 113, 2130
Sneden, C. 1973, ApJ, 184, 839
Garcia-Lario, P., Manchado, A., Pottasch, S. R., Suso, J., & Olling, R.
Soderblom, D. R., Henry, T. J., Shetrone, M. D., Jones, B. F., & Saar, S. H.
1990, A&AS, 82, 497
1996, ApJ, 460, 984 (Paper I)
Gregorio-Hetem, J., Lepine, J. R. D., Quast, G. R., Torres, C. A. O., & de la
Sylvester, R. J., & Skinner, C. J. 1996, MNRAS, 283, 457
Reza, R. 1992, AJ, 103, 549
Sylvester, R. J., Skinner, C. J., Barlow, M. J., & Mannings, V. 1996,
Herbig, G. H. 1973, ApJ, 182, 129
MNRAS, 279, 915
ÈÈÈ. 1978, in Problems of Physics and Evolution of the Universe, ed.
Thevenin, F. 1990, A&AS, 82, 179
L. V. Mirzoyan (Yerevan : Armenian Acad. Sci.), 171
Torres, G., Stefanik, R. P., Latham, D. W., & Mazeh, T. 1995, ApJ, 452, 870
Holtzman, J., et al. 1995, PASP, 107, 156
Walter, F. M., Vrba, F. J., Mathieu, R. D., Brown, A., & Myers, P. C. 1994,
Jeffries, R. D., Buckley, D. A. H., James, D. J., & Stau†er, J. R. 1996,
AJ, 107, 692
MNRAS, 281, 1001
Wichmann, R., Sterzik, M., Krautter, J., Metanomski, A., & Voges, W.
Johnson, D. R. H., & Soderblom, D. R. 1987, AJ, 93, 864
1997, A&A, 326, 211
Jones, B. F., Shetrone, M., Fischer, D., & Soderblom, D. R. 1996, AJ, 112,
Zuckerman, B., & Becklin, E. E. 1993, ApJ, 406, L25
186